The present invention relates to an I/O device switching method, and more particularly to an I/O device switching method that enables selective switching of plural I/O devices connected to a server device having an I/O switch, and a computer system that includes a management server that performs the switching control.
There are cases where a server device is connected with plural types of external devices such as network devices and storage devices that respectively have different protocols. Therefore, the server device is provided with plural I/O devices that process different protocols and make connections between the server device and external devices. In the case of general server devices, the correspondence between the server devices and the I/O devices is fixed, and cannot be changed. Therefore, in environments such as enterprises' computer systems and data centers in which numerous server devices are operating, necessary I/O devices must be selected for each of the server devices. However, if the correspondence between the server devices and the I/O devices is fixed, when the usage and the like of the server devices are changed, the I/O devices must be mounted or dismounted, making operation management complicated.
An I/O switch is one technology for solving this problem. The I/O switch is a device disposed between the server devices and the above-described I/O devices, and by controlling the configuration of the I/O switch, the I/O devices allocated to the server devices can be flexibly changed. Since plural server devices can be connected to the I/O switch, I/O devices can be flexibly allocated to the plural server devices, so that the complexity of operation management can be lessened.
A method of managing input/output interface modules is disclosed in Japanese Published Unexamined Patent Application No. 301488/2005. According to this method, a management server has a management table that manages specifications and connection states of input/output interfaces, and for an input-output device connection command from an integrated management server, consults the management table to select an input/output interface matching conditions and change its connection.
According to the related art, although the allocation of I/O devices to a server device is made flexible by the I/O switch, measures for reliability become complicated. For example, in the case of a server device having no I/O switch, in a multiplex configuration environment in which plural servers are used to achieve high reliability, when a failure occurs in any server device and it is switched to another, the server device and I/O devices included in it together are replaced. In this case, when a failure occurs in any one of in server devices, processing can be recovered by changing the server device regardless of the location of the failure. In short, since the entire server device including I/O devices is used as a switching unit, a switching method at failure can be simplified although the switching unit is large.
However, in the case of a server device including an I/O switch, since the server device is separated from I/O devices, it is necessary to correctly recognize the location of a failure and perform failure recovery suited for it. When a failure occurs in the server device, recovery from the failure can be made by changing the server device like a conventional method. However, when a failure occurs in an I/O device connected to the I/O switch, since processing of the server device is halted due to the failure in the I/O device, it appears that a failure occurs in the server device. Therefore, even if only the server device is switched using a conventional method, if the failure in the I/O device is not eliminated, recovery from the failure is impossible. For example, for recovery from a failure in an I/O device, the failed I/O device must be replaced by an I/O device of the same type as the failed I/O device. This means that a server device including an I/O switch cannot be recovered simply by detecting a failure in the server device differently from a conventional method. Therefore, by determining whether a server device fails or an I/O device fails, failure recovery suited for each case must be performed.